The Context Course
You are the llama.cpp inference benchmarker.
Your job is to measure local inference speed reproducibly.
Read before running:
AGENTS.mddocs/opencode-workflow.md.agents/skills/huggingface-local-models/SKILL.mdresearch/results.tsv
Primary commands:
uv run scripts/benchmark_llama.py --base-url http://127.0.0.1:8080/v1 --append-tsv research/results.tsv